Biodiesel production from third-generation feedstock: process parameter modelling and optimisation using RSM-ANN approach

This study employed response surface methodology (RSM) coupled with a central composite design (CCD) approach to ascertain the optimal conditions for biodiesel production from Neochloris oleoabundans microalgae oil. Four key process variables, including the methanol-to-oil molar ratio, catalyst concentration, reaction time, and temperature, were investigated across five levels to develop an L30 orthogonal array for experimentation. An artificial neural network (ANN)-based prediction model was developed using the experimentally obtained data, yielding high accuracy with mean square error (MSE) values of 0.019, 2.4327, and 0.8269 and coefficient of determination (R2) values of 0.9996, 0.9796, and 0.9890 for training, validation, and testing sets, respectively, indicating robust predictive capability. The optimisation analysis reveals a biodiesel yield of 94.94% under optimised conditions: 6.92:1 molar ratio, 1.22% catalyst concentration, 64.36 min reaction time, and 56.46 degrees C temperature. Experimental validation confirmed the reliability of the optimisation results, demonstrating a marginal error of 2%. [Received: November 25, 2022; Accepted: April 30, 2024]